Motion artifact removal by time domain projection

Abstract

An algorithm for removing motion artifacts from the PPG signal in the time domain to determine heart rate is disclosed. A device for determining a heart rate of a user can include a heart rate sensor configured to generate heart rate signals when positioned on or adjacent to a user's skin, an accelerometer configured to generate one or more acceleration signals, and processing circuitry configured to remove, in a time domain, motion artifacts from the heart rate signals based on the acceleration signals. In some examples, the removal of motion artifacts can also be based on mean-centered, variance-scaled integrated acceleration signals. In some examples, the processing circuitry can be configured to remove motion artifacts using a least squares algorithm to identify a best representation of acceleration and integrated acceleration signals in the heart rate signals.

The invention claimed is: 
     
       1. A device for determining a heart rate of a user, comprising:
 a sensor configured to generate first signals when positioned on or adjacent to a skin of the user; 
 an accelerometer configured to generate acceleration signals, wherein each of the acceleration signals comprises a signal stream of acceleration measurements of one corresponding accelerometer axis; and 
 processing circuitry capable of:
 generating integrated acceleration signals from the acceleration signals, wherein each of the integrated acceleration signals comprises a cumulative sum of the acceleration measurements of the one corresponding accelerometer axis of one of the acceleration signals; 
 removing, in a time domain, motion artifacts from the first signals using the acceleration signals and the integrated acceleration signals, wherein use of the integrated acceleration signals identifies time-delayed correlations between the first signals and the acceleration signals to account for timing mismatch between the first signals and the acceleration signals; and 
 determining the heart rate based on the first signals having the motion artifacts removed therefrom. 
 
 
     
     
       2. The device of  claim 1 , wherein the integrated acceleration signals are mean-centered and scaled to have a variance matching the acceleration signals used to generate the corresponding integrated acceleration signals. 
     
     
       3. The device of  claim 1 , wherein removing the motion artifacts comprises using a least squares algorithm to identify a best representation of the acceleration signals and the integrated acceleration signals in the first signals. 
     
     
       4. The device of  claim 3 , wherein the processing circuitry is further capable of:
 projecting the acceleration and integrated acceleration signals; and 
 generating residual first signals based on the projected acceleration and integrated acceleration signals and the first signals, wherein the residual first signals are used to determine the heart rate. 
 
     
     
       5. The device of  claim 3 , wherein the processing circuitry is further capable of:
 generating latent variables for the acceleration and integrated acceleration signals; 
 projecting one or more components of the generated latent variables; and 
 generating residual first signals based on the projected one or more components of the generated latent variables and the first signals, wherein the residual first signals are used to determine the heart rate. 
 
     
     
       6. The device of  claim 3 , wherein the processing circuitry is further capable of:
 generating first latent variables for the acceleration and integrated acceleration signals during a first iteration; 
 projecting one or more components of the first latent variables generated during the first iteration; 
 generating first residual first signals based on the projected one or more components of the first latent variables generated during the first iteration and the first signals; 
 deflating the acceleration and integrated acceleration signals based on the projected one or more components of the first latent variables generated during the first iteration; 
 generating second latent variables for the deflated acceleration and integrated acceleration signals during a second iteration; 
 projecting one or more components of the second latent variables generated during the second iteration; and 
 generating second residual first signals based on the projected one or more components of the second latent variables generated during the second iteration and the first residual first signals, wherein the second residual first signals are used to determine the heart rate. 
 
     
     
       7. The device of  claim 1 , wherein removing the motion artifacts comprises removing the motion artifacts from a plurality of overlapping time intervals to generate a plurality of overlapping residual first signals; and
 generating a final time domain first signal by averaging the plurality of overlapping residual first signals for the time intervals, wherein the final time domain first signal is used to determine the heart rate. 
 
     
     
       8. The device of  claim 1 , wherein determining the heart rate comprises determining at least one of an instantaneous heart rate, heart rate variability or average heart rate based on the first signals after removing the motion artifacts. 
     
     
       9. A method executed by processing circuitry for determining a heart rate of a user, the method comprising:
 receiving first signals generated by a sensor when positioned on or adjacent to a skin of the user; 
 receiving acceleration signals generated by an accelerometer, wherein each of the acceleration signals comprises a signal stream of acceleration measurements of one corresponding accelerometer axis; 
 generating integrated acceleration signals from the acceleration signals, wherein each of the integrated acceleration signals comprises a cumulative sum of the acceleration measurements of the one corresponding accelerometer axis of one of the acceleration signals; 
 removing, in a time domain, motion artifacts from the first signals using the acceleration signals and the integrated acceleration signals, wherein use of the integrated acceleration signals identifies time-delayed correlations between the first signals and the acceleration signals to account for timing mismatch between the first signals and the acceleration signals; and 
 determining the heart rate based on the first signals having the motion artifacts removed therefrom. 
 
     
     
       10. The method of  claim 9 , further comprising:
 projecting the acceleration and integrated acceleration signals; and 
 generating residual first signals based on the projected acceleration and integrated acceleration signals and the first signals, wherein the residual first signals are used to determine the heart rate. 
 
     
     
       11. The method of  claim 9 , further comprising:
 generating latent variables for the acceleration and integrated acceleration signals; 
 projecting one or more components of the generated latent variables; and 
 generating residual first signals based on the projected one or more components of the generated latent variables and the first signals, wherein the residual first signals are used to determine the heart rate. 
 
     
     
       12. The method of  claim 9 , further comprising:
 generating first latent variables for the acceleration and integrated acceleration signals during a first iteration; 
 projecting one or more components of the first latent variables generated during the first iteration; 
 generating first residual first signals based on the projected one or more components of the first latent variables generated during the first iteration and the first signals; 
 deflating the acceleration and integrated acceleration signals based on the projected one or more components of the first latent variables generated during the first iteration; 
 generating second latent variables for the deflated acceleration and integrated acceleration signals during a second iteration; 
 projecting one or more components of the second latent variables generated during the second iteration; and 
 generating second residual first signals based on the projected one or more components of the second latent variables generated during the second iteration and the first residual first signals, wherein the second residual first signals are used to determine the heart rate. 
 
     
     
       13. The method of  claim 9 , wherein:
 the first signals and acceleration signals from a plurality of overlapping time intervals are used to generate a plurality of overlapping residual first signals; and 
 determining the heart rate comprises generating a final time domain first signal by averaging the plurality of overlapping residual first signals for the time intervals. 
 
     
     
       14. The method of  claim 9 , further comprising measuring a time separation between two peaks in the first signals, wherein determining the heart rate comprises determining an instantaneous heart rate based on the time separation between the two peaks. 
     
     
       15. The method of  claim 9 , further comprising:
 measuring a time separation between a plurality of peaks in the first signals; 
 determining instantaneous heart rates based on the time separation between adjacent peaks of the plurality of peaks; and 
 determining heart rate variability based on the determined instantaneous heart rates. 
 
     
     
       16. The method of  claim 9 , further comprising measuring a number of peaks in the first signals during a time interval, wherein determining the heart rate comprises determining an average heart rate based on the number of peaks in the first signals during the time interval. 
     
     
       17. The method of  claim 9 , wherein the integrated acceleration signals are mean-centered and scaled to have a variance matching the acceleration signals used to generate the corresponding integrated acceleration signals. 
     
     
       18. The method of  claim 9 , wherein removing the motion artifacts comprises using a least squares algorithm to identify a best representation of the acceleration signals and the integrated acceleration signals in the first signals. 
     
     
       19. A non-transitory computer readable storage medium, the non-transitory computer readable storage medium containing instructions that, when executed, perform a method for operating an electronic device, the electronic device including a processor, the method comprising:
 receiving first signals generated by a sensor when positioned on or adjacent to a user's skin; 
 receiving acceleration signals generated by an accelerometer, wherein each of the acceleration signals comprises a signal stream of acceleration measurements of one corresponding accelerometer axis; 
 generating integrated acceleration signals from the acceleration signals, wherein each of the integrated acceleration signals comprises a cumulative sum of the acceleration measurements of the one corresponding accelerometer axis of one of the acceleration signals; 
 removing, in a time domain, motion artifacts from the first signals using the acceleration signals and the integrated acceleration signals, wherein use of the integrated acceleration signals identifies time-delayed correlations between the first signals and the acceleration signals to account for timing mismatch between the first signals and the acceleration signals; and 
 determining a heart rate based on the first signals having the motion artifacts removed therefrom. 
 
     
     
       20. The non-transitory computer readable storage medium of  claim 19 , the method further comprising:
 projecting the acceleration and integrated acceleration signals; and 
 generating residual first signals based on the projected acceleration and integrated acceleration signals and the first signals, wherein the residual first signals are used to determine the heart rate. 
 
     
     
       21. The non-transitory computer readable storage medium of  claim 19 , the method further comprising:
 generating latent variables for the acceleration and integrated acceleration signals; 
 projecting one or more components of the generated latent variables; and 
 generating residual first signals based on the projected one or more components of the generated latent variables and the first signals, wherein the residual first signals are used to determine the heart rate. 
 
     
     
       22. The non-transitory computer readable storage medium of  claim 19 , the method further comprising:
 generating first latent variables for the acceleration and integrated acceleration signals during a first iteration; 
 projecting one or more components of the first latent variables generated during the first iteration; 
 generating first residual first signals based on the projected one or more components of the first latent variables generated during the first iteration and the first signals; 
 deflating the acceleration and integrated acceleration signals based on the projected one or more components of the first latent variables generated during the first iteration; 
 generating second latent variables for the deflated acceleration and integrated acceleration signals during a second iteration; 
 projecting one or more components of the second latent variables generated during the second iteration; and 
 generating second residual first signals based on the projected one or more components of the second latent variables generated during the second iteration and the first residual first signals, wherein the second residual first signals are used to determine the heart rate. 
 
     
     
       23. The non-transitory computer readable storage medium of  claim 19 , wherein:
 the first signals and acceleration signals from a plurality of overlapping time intervals are used to generate a plurality of overlapping residual first signals; and 
 determining the heart rate comprises generating a final time domain first signal by averaging the plurality of overlapping residual first signals for the time intervals. 
 
     
     
       24. The non-transitory computer readable storage medium of  claim 19 , wherein determining the heart rate comprises determining at least one of an instantaneous heart rate, heart rate variability or average heart rate based on the first signals after removing the motion artifacts. 
     
     
       25. The non-transitory computer readable storage medium of  claim 19 , wherein the integrated acceleration signals are mean-centered and scaled to have a variance matching the acceleration signals used to generate the corresponding integrated acceleration signals.